Try Ditto free

Custom AI Agent Development

We build the AI agents your business needs

We develop AI agents that handle the specialised jobs in your business. We train them on your processes and give them persistent, private memory that is never used to train AI models.

Get in touchTry Ditto free
Jobs we build agents for
You tell us

“Every new client should get a welcome pack within a day.”

We build
Onboarding agentEvery morning, 8am
MemoryClient filesOnboarding checklist
Tools
First output
✓Welcome pack ready
  • Welcome email drafted
  • Checklist filled from HubSpot
  • Missing documents listed

What custom AI agents change in your business

What changes once agents are built around the jobs your team repeats.

Scale without adding headcount

Agents take on the jobs your team repeats, every day or on a schedule.

No more manual handoffs

Agents pass work to each other, so nobody copies results from one app to the next.

Knowledge stays when staff leave

What your team and agents save to team memory stays with the business.

Every agent starts informed

Each agent starts from what your business already knows, not a blank prompt.

Our AI agent development services

Five ways we put AI agents to work, from a single agent to a connected team.

Custom AI agent development

Agents built for the specific jobs your teams repeat, from client onboarding to the weekly report.

Multi-agent system design

A team of agents with a lead agent that hands out the work and agents that pass it on.

AI integration services

Agents connected to the apps your business runs on, and to your own systems.

Workflow automation agents

Jobs described in plain words that run on demand or on a schedule.

Knowledge agents built on your company memory

Agents that answer from what your business has saved and name their sources.

Tell us the job you want an agent forEvery new client should get a welcome pack within a day. Tell us the job

How we build your custom AI agents

Eight steps, from the first call to agents that keep getting better.

  1. 01

    Discovery and job mapping

    We learn how your business works and map the jobs your team repeats.

    This step includes:

    • The jobs each team repeats each week
    • The apps and systems each job touches
    • Which jobs agents should take on first
  2. 02

    Spec and quote

    You get a written spec and a quote sized to it, before any build starts.

    This step includes:

    • Scope, agents and workflows
    • Deployment and who can see what
    • Success criteria for each agent
  3. 03

    Connect your apps and set access

    We connect the apps each team uses and set access from each app's own permissions.

    This step includes:

    • Native connectors and MCP connections
    • Deeper connections to your own systems
    • Read-only until sending is switched on
  4. 04

    Set up your business memory

    Your connected apps start filling one memory, organised so each agent can start from the right part of it.

    This step includes:

    • Memory built from your apps and chats
    • A separate memory for each client where needed
    • The memory each agent starts with
  5. 05

    Shape and train each agent

    We train each agent on your processes through its instructions, notes, memory and tools. We don't train or fine-tune AI models on your data.

    This step includes:

    • Instructions and notes from the people who do the job
    • Tools and app actions for each agent
    • A lead agent to hand out work in a team
  6. 06

    Build and test the workflows

    We build the jobs each agent runs and try them with your real data before they go live.

    This step includes:

    • Workflows that run on demand or on a schedule
    • Ready-made workflows where one fits
    • Trial runs that hold back sends and changes
  7. 07

    Train your team and go live

    Your team learns how to work with the agents, and the agents start on the jobs in the spec.

    This step includes:

    • How to ask Ditto and review its work
    • How to see, edit and delete memory
    • When to switch sending on for each app
  8. 08

    Optimise and improve

    We stay on after launch to keep your agents working as your business changes, and every job adds to what they know.

    This step includes:

    • Support, maintenance and fixes
    • Instructions refined as your processes change
    • New features as your business grows

Key components of a custom AI agent built on Ditto

The parts we set up for each agent we build.

  • Instructions and notes

    What the agent does, how it works and the rules it follows. We write them with the people who do the job now.

    Written with your team

  • Memory context

    The memories the agent starts with, such as one client's files or past decisions. Each new agent can build on what was already saved.

    Starts from saved knowledge

  • Tools and app actions

    The apps and tools the agent can use. It reads and acts through each app's own actions.

    Acts in your apps

  • Workflows and schedules

    The jobs the agent runs, step by step. A workflow runs when you start it or on the schedule you set.

    Runs on your schedule

  • Lead agent and messaging

    In a team, a lead agent hands tasks to the others. Agents message each other to pass work on.

    Agents work as a team

  • Permissions and access

    Agents work inside the access your team already has in each app.

    Inside your access rules

Every agent runs on Ditto's memory of your business

Ditto is the memory under every agent we build. Connect your apps once and every agent works from the same record of your business.

  • Fills from your apps

    Email, chat, internal documents and records from your connected apps are saved together, along with your team's chats with Ditto.

  • Each agent starts informed

    Choose the memory each agent starts with, such as everything saved about one client.

  • Shared across agents

    What one agent saves, every agent can use on its next job.

  • Stays with the business

    Team memory stays when a person leaves.

For technical readers: Ditto organises memories into a knowledge graph of subjects. When an agent works, retrieval finds relevant memories by semantic similarity, ranks them and adds the best matches to the model's context window.

Use cases of AI agents across industries

Examples of agents we can build for operations, client and customer teams in each industry.

Support reply agent

Agent
Draft ready

Finds the order status in Shopify and any supplier email that moved the date, then drafts responses to customers for your support team.

Restock update agent

Agent
Draft ready

Matches restock dates from supplier and 3PL emails to each product and drafts the product page and customer updates.

See Ditto for ecommerce and retail

Security and privacy for your AI agents

You decide what your agents can do, and nothing they read is used to train AI models.

Your business data

  • Not used to train modelsNothing you connect or say in Ditto is used to train AI models.
  • Models that keep no copyChoose models whose provider keeps no copy of your data.
  • One client never sees anotherKeep a separate memory for each client or project.
  • See and delete any memoryOpen any memory to see where it came from, edit or delete it, or export all of it as a signed Ditto Passport.

Your control

  • Read-only to startDitto starts read-only in each app. You switch on sending app by app.
  • Approval before it goes outSends and changes wait for your approval unless you let a workflow run them.
  • Permissions checkedDitto checks each app's permissions before it shares information.
  • Disconnect any appDisconnecting an app stops Ditto reading it.

Ditto's memory harness and integrations are open source.How Ditto handles your data

Two ways to get custom AI agents

We build it

Our team builds your agents

We run discovery, write the spec, build your agents and workflows, and train your team. Then we stay on to support and improve them.

  • Discovery and a written spec
  • Agents and workflows built for you
  • Team training and handover
  • Ongoing support and improvements
Get in touch
You build it

Build agents yourself in Ditto

Create an agent in one click in the Ditto app, give it instructions and the memory it starts with, then build workflows for it.

  • Create agents in one click
  • Write instructions and choose each agent's memory
  • Build workflows in plain words
  • Start on a self-serve plan
See AI agent teams

What custom AI agent development costs

Each build is quoted to its written spec.

Setup fee

Customone-off

  • ✓Discovery and a written spec
  • ✓Agents, workflows and app connections
  • ✓Team training and handover
Monthly retainer

Customa month

  • ✓Hosting and support
  • ✓Maintenance and fixes
  • ✓New features as your business grows

Get in touchSee bespoke business plans

We connect your agents to the apps your teams work in. Over 1,400 integrations.

See all app integrations

Custom AI agent development FAQ

How much does custom AI agent development cost?

A build has two parts: a one-off setup fee for the spec, the build and deployment, and a monthly retainer for hosting, support, maintenance and new features. Both are quoted to your written spec.

How do you train a custom AI agent for my business?

We shape each agent with instructions and notes written with the people who do the job, the memory it starts with and the tools it can use. Ditto uses models from providers such as OpenAI, Anthropic and Google. It does not train or fine-tune them on your data.

What is the difference between an AI agent and a chatbot?

A chatbot answers customer questions in one chat window. An AI agent built on Ditto works across your connected apps: it reads email, documents and records, prepares drafts and summaries, and runs workflows on a schedule, starting from your business memory each time.

Can the AI agents connect to our existing systems?

Yes. Agents work in apps such as Gmail, Slack, HubSpot and QuickBooks through each app's own actions, and through MCP. As part of a build we can add deeper connections to the tools your business runs on, beyond the standard connectors.

How do you keep our data private and secure?

Ditto checks each app's permissions before it shares information and starts read-only in each app. Nothing is used to train AI models. You can keep a separate memory for each client, choose models whose provider keeps no copy, and delete any memory.

Is our data used to train AI models?

No. Nothing you connect or say in Ditto is used to train AI models. Ditto doesn't train its own foundation models. It uses models from providers such as OpenAI, Anthropic and Google.

Can several AI agents work together?

Yes. We can build a team with a lead agent that hands out the work. The agents message each other, and what one agent saves, the others can use.

What happens after the agents go live?

We stay on to support your agents, fix issues and add new features as your business changes. Your agents keep saving what they learn, so later jobs start with more of your business's history. This is covered by the monthly retainer.

Do you have AI agent templates?

Ditto has ready-made workflows you can start from, and it can turn a skill written for Claude or another agent into a workflow. In a custom build we start from those where they fit and build the rest for your business.

Can we start on a self-serve plan first?

Yes. Start on a self-serve plan and build agents yourself in the Ditto app. Get in touch when you want our team to build agents for your business.

Do you build AI coding agents?

Ditto's coding agent features are in early access. Our coding agent memory page shows how coding agents such as Claude Code and Cursor share Ditto's memory.

Get 20% off your first month of Ditto, and get more out of it every week.